Bandwidth Extension Module for Speech Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech signal processing technologies fail to effectively enhance the perceived quality of band-limited speech signals, particularly in public telephone networks, due to the absence of high-frequency components, which leads to muffled sounds and difficulties in distinguishing certain phonemes.
Innovation Solution
A method and apparatus that artificially extend the bandwidth of lowband speech signals through band-pass filtering, pitch-synchronous modulation, and combination with a highband speech signal component to produce a bandwidth-extended speech signal, using a bandwidth extension module that includes a band-pass filter, carrier frequency modulators, and a synthesis filter.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If speech signals are band-limited to 300-3400 Hz for digitization at 8 kilosamples per second, then bandwidth requirements are reduced and intelligibility is retained, but perceived naturalness deteriorates and high-frequency phoneme distinction capability is lost
Solution Approach 1:
The patent applies preliminary action by pre-processing the lowband speech signal through band-pass filtering to extract relevant spectral components before modulation. This preparation enables the subsequent pitch-synchronous modulation to effectively generate highband content that mimics the original wideband speech characteristics, thereby improving perceived quality without requiring actual wideband transmission
Solution Approach 2:
The patent changes the frequency parameter of the speech signal by modulating the band-pass filtered signal at pitch-synchronous intervals. This parameter transformation shifts energy from the lowband range to the highband range, artificially creating high-frequency content that was originally absent, thus resolving the contradiction between limited bandwidth and perceived quality
2Manufacturing precision
If existing spectral-folding techniques are used to generate highband excitation signals, then bandwidth extension is achieved, but artifacts appear in the resulting wideband speech when the band-limited speech contains nasal phonemes
Solution Approach 1:
The patent extracts only the essential spectral envelope information from the lowband speech signal through band-pass filtering, separating it from the problematic excitation components that cause artifacts in nasal phonemes. By taking out only the necessary spectral shape information and regenerating the excitation through pitch-synchronous modulation, the method avoids the artifact-generating mechanisms of spectral-folding techniques
Solution Approach 2:
The patent introduces pitch-synchronous modulation as an intermediary process between the lowband speech signal and the highband excitation signal generation. This intermediary mechanism transforms the band-pass filtered signal in a controlled manner that preserves speech naturalness while avoiding the direct spectral manipulation that causes artifacts in conventional spectral-folding methods
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Improves the perceived quality of speech by populating the highband range with frequency content, enhancing the ability to distinguish phonemes and providing a high-quality speech experience similar to wideband-capable devices.
Implementation Method 1
band-pass filtering the lowband speech signal to obtain a band-pass signal
Implementation Method 2
pitch-synchronously modulating the band-pass signal about at least one carrier frequency to obtain a highband speech signal component
Implementation Method 3
The highband excitation signal is passed through a linear prediction synthesis filter defined by the highband spectral parameters in order to generate the spectral content in the highband range
Data Source
AI summary
A bandwidth extension module, and an associated method and computer-readable medium, suitable for use in artificially extending the bandwidth of a lowband speech signal. The bandwidth extension module comprises a band-pass filter configured to produce a band-pass signal from the lowband speech signal; at least one carrier frequency modulator, each carrier frequency modulator configured to pitch-synchronously modulate the band-pass signal about a respective carrier frequency, the at least one carrier frequency modulator collectively producing a highband speech signal component; a synthesis filter configured to determine a highband speech signal based on the highband speech signal component; and a summation module configured to combine the lowband speech signal with the highband speech signal to obtain a bandwidth-extended speech signal.